In contrast to many of the AI applications known today, which primarily analyse and provide recommendations, agentic AI can act independently. It not only recognises patterns in data, but also plans, coordinates, and manages entire process chains within defined limits. It thus marks the transition of AI from a tool to an active digital assistant.
Guest article Agentic AI: next important step for efficient health care
The co-authors of the focus paper, Heyo Kroemer and Alexander Meyer, on “Agent-Based AI in the Healthcare System: Opportunities and Challenges”
- Generative AI
- Large Language Models
- Healthcare
- Health Policy
Health care in Germany is under pressure. The shortage of skilled workers is exacerbating difficulties in hospitals, treatment centres, care homes, and doctors’ practices, the population is ageing, bureaucracy slows progress, and all while medical care is becoming more complex. While political debates mainly revolve around questions of funding and structural reforms, a technology is developing in the background that has the potential to change many of these challenges fundamentally: agentic artificial intelligence.
Related Links
Expert on the topic Professor Dr Heyo K. Kroemer ▸
- Physiology and Pharmacology/Toxicology
- Election year 2018
An example from medicine illustrates the difference: If a conventional AI recognises a heart attack based on ECG data, it provides a diagnostic probability. Agentic AI, however, could also alert the responsible treatment team, check available resources, coordinate diagnostic steps, prepare documentation, and monitor the ongoing treatment process. People remain responsible for the actual medical care, but the organisation of processes becomes much more efficient.
This is where the great potential lies. A major part of the labour involved in health care arises not from the actual treatment of patients, but rather from tasks relating to documentation and administration. Agentic AI could largely automate these processes. The burden on doctors and care workers would be lightened and they could concentrate more on their core task of caring for people.
Expert on the topic Professor Dr Alexander Meyer ▸
- Institute for Artificial Intelligence in Medicine, Charité
- Berlin
The technology has the potential not only to boost efficiency, but also to improve the quality of treatment. Agentic AI is capable of combining large amounts of structured and unstructured data, adhering to guidelines, and proposing individual treatment plans. This creates new possibilities for personalised medicine, particularly in the case of chronic diseases or complex therapies. Digital assistants could offer continuous support to patients, while medical teams would be alerted early to risks or complications.
There is also a strategic aspect involved. Agentic AI is developing into a global key technology. At the moment, above all US companies dominate the market for high performance AI systems. Europe and Germany have access to excellent research, but still far too rarely manage to transfer scientific insight into products that can be developed to the point where they enter the market. There is thus a risk of growing technological dependence, in particular in health care.
What is at stake here is not just innovation, but also digital sovereignty. Germany ought to combine its strong medical research with expertise from computer science and data science and support specific AI systems that are tailored for use in hospitals and medical practices in particular specialist areas. Regional centres of excellence, close cooperation between research and industry, as well as investment in digital infrastructure can lay the foundation for this.
While agentic AI offers great opportunities, we also cannot ignore the risks. As AI systems increasingly make independent decisions and initiate processes, new questions arise, particularly with respect to responsibility and control. Who is responsible if decisions go wrong? How can sensitive health data be protected? And how to ensure that medical staff do not lose important skills if tasks are constantly delegated to AI? Cybersecurity is becoming more important: The more deeply AI systems are embedded in clinical processes, the more attractive they become as potential targets for attack. People will thus only trust the technology if security, transparency, and ultimate human accountability are given consideration from the very start.
The decisive question is therefore not whether agentic AI will be part of our health system. Technological development means this reality is already foreseeable. The true challenge is in creating the appropriate framework conditions: innovation-friendly legislation, reliable data-protection standards, high-performance digital infrastructure, and active support for European solutions.
Agentic AI is not a scenario of the distant future. It has the potential to be one of the most important means of delivering high-quality, efficient, and sustainable healthcare despite scarce resources. Now is the time to create the conditions to make this a reality.